I know one of them is that they can easily fail catastrophically;
I meant the
reasons for them to fail catastrophically. If they "just failed", no one would use them. They are reliable if used correctly. They fail only for specific reasons.
The reasons are:
1) overvoltage; mind the industry-accepted bogus* ratings of voltage (this method of rating is not seen on
any other component type AFAIK) so that the designer needs to read from
a separate official document that the actual voltage rating is only 50-70% of the datasheet value,
then add any derating for spikes or reliability.
*) Why I call this bogus? A comparison to make the point:
* an aluminium elcap datasheet truthfully tells you that if you operate the capacitor at all maximum ratings, the lifetime will be, for example, 2000 hours. A separate appnote tells you how to make this value, which is often already acceptable as is, longer if you need to. You apply normal engineering derating to compensate for any uncertainty
you have about
your design.
* the tantalum datasheet just tells you the voltage, capacitance, hopefully ripple current ratings, no information about the lifetime. Then you have to know there's a separate document (which the datasheet does not refer to), which tells you that the voltage ratings on the datasheet are for the operating life of, for example,
2 hours, completely impractical for any design. Then they go on and tell you to derate by 40%. An absolute joke. Why do they do this? Tradition, it's been like this for a long time. If any manufacturer started to rate truthfully, their products would seem inferior. Serious customers are not complaining, it's just the right amount of black magic which makes professional design engineers proud of their "secret" information. Yes, I can design with tantalums safely.
2) strict current ratings, often preventing the usage as hot-pluggable input capacitors, or as switch mode converter main output capacitors. Even a single pulse event can cause a catastrophic failure.
And it indeed tends to happen that when someone knows 1), they forget about 2), or vice versa. And it's always designer's fault, of course.
Tantalums are not very suitable for this case, for two reasons:
1) They seem to want you to provide low ESR (<0.1 ohm); tantalums exist and are used to provide
high ESR (~ohms),
2) I can see how a DC voltage could be hot plugged to the input of these filters, causing overcurrent damage to the tantalums.
Tantalums are usable when you need stable, high ESR values, or large stable capacitance values in low-current (all peaks included) circuits, like after a voltage reference or a small, current-limited linear regulator, especially when they are not stable with MLCC output caps.